The white paper (http://senseable.mit.edu/urbanvillages/img/manuscript.pdf http://senseable.mit.edu/urbanvillages/img/manuscript.pdf) gives the assumptions to e

The white paper (http://senseable.mit.edu/urbanvillages/img/manuscript.pdf http://senseable.mit.edu/urbanvillages/img/manuscript.pdf) gives the assumptions to explain why call data is representative: "Mobile phone communication data are not necessarily a direct representation of the underlying social network. For instance, two individuals may maintain a strong tie through face-to-face interactions or other means of communication, without relying on regular phone calls [23]. Nevertheless, despite such a po- tential bias, a recent comparison with a questionnaire-based survey has shown that mobile phone communication data are, in general, a reliable proxy for the strength of individual- based social interactions [25]. Moreover, even if two subscribers maintain a close relationship and usually communicate via other means, it seems reasonable to assume that both individ- uals have called each other at least once during the relatively long observation period of 15 months, thus reducing the chance of missing such relationships in our network [21, 26, 27]"